Assessment and correction of three satellite rainfall estimate products for improving flood prevention in French Guiana

The French Guiana (80 000 km(2)) is highly vulnerable to flooding during the rainy season but the hydrological prevision is limited because the region cannot be cover by a dense network of rain gauges. Meteorological satellites could be an alternative for the measurement of precipitation. The objective of this paper was to evaluate and improve the accuracy of daily satellite rainfall estimates (SRE) throughout the French Guiana between April 2015 and March 2016. Validation data were composed by 70 rain gauges managed by France and Suriname. Three satellite-based rainfall estimates have been tested: TRMM-TMPA 3B42 (Tropical Rainfall Measuring Mission Multi-Satellite Precipitation Analysis) V7, IMERG (Integrated Multi-satellitE Retrievals) for GPM (Global Precipitation Measurement) and STAR Satellite rainfall estimates Hydro-Estimator (HE). Better SRE were obtained by GPM with a clearly higher probability of detection of rainy days (>70%). During the rainy season, biases remained important and SRE appeared inaccurate for the monitoring and forecasting of floods. Biases correction methods were applied, and the additive correction methods by interpolation of biases (ADD_IDW) obtained the better performance (absolute biases <8 mm day(-1); RMSE <12 mm day(-1)) for each satellite products. This simple method proved to be very effective to reduce biases close to 0 throughout the year. After ADD_IDW correction, performance levels of TRMM, GPM, and HE products were relatively close and these three satellite products could be implemented into cascade chains in operational framework ensuring the provision of corrected SRE in real time and thus guarantee a reinforced hydrological monitoring in French Guiana.